Dave, after having some bifurcated springs fail I have evolved a different way to make single leaf springs.  I soft solder the spring between the chops, and make the latter from a pair of profiled out 1.6mm brass plates as shown above – sorry this is a snip of the CAM profile, the grid has 0.5mm squares.  The plates are rectangular and have an aperture milled in the centre of each (which of course ends up with radiused corners) and the circles are spiral drilled holes.  The spring shim strip, as long as the plates are wide and slightly narrower than the aperture, is carefully laid across the lower plate so it is symmetrical and the upper plate clamped on with nuts+bolts at each corner.  Then I solder the assembly having masked the spring on each side in the aperture.  You might rivet rather than solder through the holes on the centre line or extra holes. You would want to put rivets through the actual spring to maximise the clamping force. Finally once the spring is securely fixed mill away the sides of the plates including the corner screw holes.  This process gets the spring fully aligned and everything square.

I don’t find much benefit of double leaf springs, they are harder to make and difficult to align.  The loss in the spring is likely to be minimal compared to other losses.  Double springs are useful to help suppress pendulum twist but yours are so close together they probably wouldn’t help much.

The length of brass is minimal, it’s easy to machine and solder, probably better than steel.  If you don’t solder you might make sure the chop clamping surfaces are really flat so the clamping points are right at the corners of the chops.  Make sure there’s no burr on the corner though.

The spring will bend where the physics tells it to, which is not far below the edge of the chop (if that is sharp and square).  But it will be tangential to the chops (the physics says), so no benefit from rounding the chops’ corners.

In this design the spring length between the chops is 6mm – no point in making it any longer, just gets harder to make and more thermal expansion.  With a heavier pendulum the effective “hinge line” moves upwards.  This went in a clock where the bob is ~250gsm on a 1/2 sec pendulum.

The last thing you want to do is etch a groove!  It will just cause stress concentration and premature fracture.

On the other hand, Matthys describes a way to make combined springs and chops by starting with thick phosphor bronze bar and milling away on each side to leave a longer section in the middle for the actual spring.  You mount the bar horizontal, and use the side of the cutter to make a long “flat”.  Then the bar is turned over and the space between the flat and a supporting angle plate filled with a hard setting filler to support the spring section as the other side is milled away.  He claims good results but I can imagine a lot of scrap on the way.  And in my experience PB is a PIG to machine!

By the way, Cousins have a wide range of suspension springs of all sorts – I did buy some double leaf springs from them once but they weren’t square.  I believe that Chris Raynerd may use ready-made Kundo springs on his Tekippe regulator – certainly some of the makers in the US are.

If the shim is clamped between 2 brass plates how does the solder get in to the non existent gap? It obviously does, so I’m missing something.

What do you use to keep it off where it’s not wanted?

On 24 November 2025 at 23:00 duncan webster 1 Said:

If the shim is clamped between 2 brass plates how does the solder get in to the non existent gap? It obviously does, so I’m missing something.

What do you use to keep it off where it’s not wanted?

Flux the surfaces before clamping. Capillary action pulls solder in.

On 25 November 2025 at 14:14 Robert Atkinson 2 Said:

Hi Dave,
The Post Office gauge is a preset type. The force set on the scale is reached as soon as the lever a.k.a pointer starts to move off it’s neutral position. difference from first move to stop is insignificant.

See:
https://www.britishtelephones.com/gec/journal/currentcomments/1936_vol6_88.pdf

Interesting document, Robert … Thanks

The patent is available on Espacenet

MichaelG.

If the losses are primarily air drag, for a given bob shape and size the losses are independent of the mass so the required impulse power doesn’t change.  A higher mass gives a higher Q of course.